Article Per Year (5 Year)
p-Index From 2020 - 2025
0.61
P-Index
This Author published in this journals
All Journal Jurnal Algoritma Jurnal Pilar Nusa Mandiri Jurnal Algoritma
Thira, Indra Jiwana
Unknown Affiliation
Computer Science & IT

Published : 3 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 3 Documents
Search

 1 
Pengenalan Alfabet Sistem Isyarat Bahasa Indonesia (SIBI) Menggunakan Convolutional Neural Network Thira, Indra Jiwana; Riana, Dwiza; Ilhami, Azriel Noer; Dwinanda, Brama Rizky Setia; Choerunisya, Hana
Jurnal Algoritma Vol 20 No 2 (2023): Jurnal Algoritma
Publisher : Institut Teknologi Garut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33364/algoritma/v.20-2.1480

Abstract

Deaf is fourth in the list of persons with disabilities in Indonesia at 7.03%. Deaf people communicate using sign language both when communicating with fellow deaf people and with normal people. The problem that arises is that few normal people master sign language, especially the Indonesian Sign System (SIBI) so that it becomes an obstacle when they have to communicate with deaf people. This study aims to classify the alphabet in SIBI except the letters J and Z with a total of 24 classes. Classification is done by comparing three CNN architectures, namely MobileNetV2, MobileNetV3Small and MobileNetV3Large to get the best model. The results showed that the MobileNetV3Small architecture produced the best model at batch size 32 and the number of epochs 30 with an accuracy of 98.81% for testing data.
IMAGE BACKGROUND PROCESSING FOR COMPARING ACCURACY VALUES OF OCR PERFORMANCE Kholifah, Desiana Nur; Nawawi, Hendri Mahmud; Thira, Indra Jiwana
Jurnal Pilar Nusa Mandiri Vol 16 No 1 (2020): Pilar Nusa Mandiri : Journal of Computing and Information System Publishing Peri
Publisher : LPPM Universitas Nusa Mandiri

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (1274.786 KB) | DOI: 10.33480/pilar.v16i1.1076

Abstract

Optical Character Recognition (OCR) is an application used to process digital text images into text. Many documents that have a background in the form of images in the visual context of the background image increase the security of documents that state authenticity, but the background image causes difficulties with OCR performance because it makes it difficult for OCR to recognize characters overwritten by background images. By removing background images can maximize OCR performance compared to document images that are still background. Using the thresholding method to eliminate background images and look for recall values, precision, and character recognition rates to determine the performance value of OCR that is used as the object of research. From eliminating the background image with thresholding, an increase in performance on the three types of OCR is used as the object of research.
Pengenalan Alfabet Sistem Isyarat Bahasa Indonesia (SIBI) Menggunakan Convolutional Neural Network Thira, Indra Jiwana; Riana, Dwiza; Ilhami, Azriel Noer; Dwinanda, Brama Rizky Setia; Choerunisya, Hana
Jurnal Algoritma Vol 20 No 2 (2023): Jurnal Algoritma
Publisher : Institut Teknologi Garut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33364/algoritma/v.20-2.1480

Abstract

Deaf is fourth in the list of persons with disabilities in Indonesia at 7.03%. Deaf people communicate using sign language both when communicating with fellow deaf people and with normal people. The problem that arises is that few normal people master sign language, especially the Indonesian Sign System (SIBI) so that it becomes an obstacle when they have to communicate with deaf people. This study aims to classify the alphabet in SIBI except the letters J and Z with a total of 24 classes. Classification is done by comparing three CNN architectures, namely MobileNetV2, MobileNetV3Small and MobileNetV3Large to get the best model. The results showed that the MobileNetV3Small architecture produced the best model at batch size 32 and the number of epochs 30 with an accuracy of 98.81% for testing data.
Co-Authors Choerunisya, Hana Dwinanda, Brama Rizky Setia Dwiza Riana Ilhami, Azriel Noer Kholifah, Desiana Nur Nawawi, Hendri Mahmud